Updated 20 September 2026: first published on Elite Agent in April 2025. The article has been refreshed as an AI basics guide to practising clear, thoughtful client conversations.
“Why are your fees higher?” “Why would I use a property manager?” “Can I have a little more time?”
These are useful conversations to practise before you need to have them for real. If you are starting with AI, a fictional role-play gives you a small, repeatable exercise: listen, answer, review and try again.
I explored this with the Sales Objection Game in my 2024 AI Powered Agents workshops. You can use that original exercise or run a practice conversation in your own AI chat.
The goal is to explain your service clearly and understand the concern. A client may still want time, more information or a different service.
The original game’s opening screen is shown below. The prompt that follows also works as a standalone practice brief.
How do I practise a client conversation with AI?
Choose one situation. For example, a fictional landlord has managed a property themselves and is unsure what a property manager would add.
List the facts you can support about your own service. That might include your agreed reporting process or the tasks included in your management agreement. Do not borrow somebody else’s experience, results or promises.
Use a fictional scenario and leave out real client details. Then try this prompt:
Help me practise a real estate client conversation using a fictional scenario.
My role and service: [DETAILS]
Verified service facts I can rely on: [FACTS]
Client’s concern: [OBJECTION]
Play the client. Ask one question at a time and wait for my answer. Do not invent laws, market evidence, results or claims about my experience.
After I answer, comment on clarity, relevance, empathy, evidence and conciseness. Quote the part of my answer you are discussing, suggest one improvement and let me try again. Flag unsupported claims instead of making them sound more persuasive.
Let the client remain unconvinced, ask for time or say no. Do not treat agreement as the only successful outcome. If you give a score, explain that it is practice feedback, not a prediction of sales ability or client behaviour.
Finish with one thing to check and one thing to practise with a colleague.What should I learn from the feedback?
In the original example, the AI gave an answer 31 out of 50. It called a reference to “new real estate rules” vague and suggested acknowledging the landlord’s experience more directly.
Those are useful points to inspect, but the score itself is not a measurement of professional ability. Nor does the chatbot’s simulated “I’ll buy” tell you what a real person will do.
The screenshot is from the original role-play. References in that example to 35 years of experience or legal changes belong to the example; check your own facts before using any similar wording.
My observation that first attempts at the 2024 workshops had not exceeded 75 per cent was an anecdote from those sessions, not a benchmark to aim for. Feedback can change with the model, prompt and conversation.
Try one better answer, then practise it aloud
For the fictional self-managing landlord, a useful opening might be: “It sounds like you’ve put a lot of work into managing it yourself. Which parts would you want help with, if any?”
That is an invitation to understand the person’s needs. Their answer should shape what you explain next.
Review the AI’s advice as critically as your own response. Did it address the actual concern? Did it suggest a claim you cannot substantiate? Is the revised answer something you would comfortably say in person?
Try the conversation with a colleague. Ask whether your explanation is clear, whether your question leaves room for an honest answer and whether the next step suits the client.
A short practice session is enough to start. Choose one improvement you can hear in your next answer.
The original version appeared in Samantha McLean’s AI Powered Agents newsletter. Read the refreshed source article on AI Powered Real Estate by Samantha McLean on Substack.